ScaleIO: When More is Better…and Less is More!

By now, you’ve no doubt heard how Software Defined Storage (SDS) and hyper-converged server SANs are reshaping the storage industry (and use case) landscape.  Market acceptance continues to grow at an accelerated pace.  There are now many established and new vendors offering hyper-converged storage appliances and SDS systems.   EMC’s own ScaleIO SDS platform continues to score mounting recognition, kudos and accolades with our growing list of customers.Hyperconverged  ScaleIO has even won-over storage techies and IT Management who were initially skeptical that ScaleIO would actually completely deliver on its value proposition and capabilities.

Why all the buzz and hype?  Customers – and the IT datacenter and storage market — get it and are embracing this new, “disruptive” technology by introducing it to their enterprise data centers or using it to create hyper-scalable virtualized infrastructures for cloud applications.  The promise and appeal of installing software on individual commodity host application servers to create a virtual storage pool (i.e., “server SAN”) from each participating server’s excess direct attached storage (DAS) without requiring additional specialized storage/fabric hardware is alluring…and almost too good to be true.  High Availability, commodity based hardware, linear performance and storage capacity scalability, “SAN-less” data storage, lower TCO, elasticity and flexibility are just a few compelling arguments in favor of hyper-converged architectures in general and SDS system in particular.

But just how does ScaleIO in particular stack up against our esteemed competitors?   Let’s take a close snap look at Nutanix and their NX Series hyper-converged storage appliances.  Nutanix, offers essentially a 2U hybrid storage server, a hypervisor (ESXi, Hyper-V, KVM and their new ‘homegrown’ Acropolis) and the Nutanix software. But what do customers really get with a Nutanix appliance?

Here’s their major value props from a 30,000 feet view…

  • A 2U server limited to a max of 20TBs of SSD and HDD raw capacity
  • Up to 4 self-contained (but pooled) nodes per chassis
  • Nutanix software that will only run on virtualized environments (VMs), not physical bare metal.
  • Data protection using simple RAM/SSD caching (locally) and distributed striping/mirroring across a Nutanix cluster
  • Hyper-converged “datacenter in a box” …with client/application, CPU, storage capacity and networking (GbE) all combined together in a single node
  • “A datacenter in a box” approach
  • Parallel I/O performance linear scalability via server SAN shared CPU processing and I/O pooling across a distributed/shared cluster
  • An appliance box optimized for VDI and virtualized (i.e., lots of Random Reads) use cases

Some good stuff for sure.  But with ScaleIO SD you get much more of the great stuff:

  • Up to 16 Petabyte of raw capacity per server of your choice, with more usable storage
  • Support from 3 to literally “thousands” of nodes – virtual or physical
  • Enterprise high availability with up to 8 separate Protection Domains supporting thousands of volumes, RAM caching and Server Fault Sets
  • Parallel I/O performance IOPS that linearly scale to over 10M IOPS from commodity hardware
  • Ability to create storage pools and performance tiers (instead of straight LFU/LRU caching)
  • Enterprise management including on demand, real-time ‘on-the-fly’ elasticity for adding or removing host servers or storage devices (and raw storage capacities)

ScaleIO PerformanceAnd I haven’t come close to covering all of ScaleIO’s additional enterprise and QoS features…that delivers so much more than what a dense Nutanix environment with its ‘noisy neighbor’ effect can offer.  Noisy neighbor? Let’s not overlook the fact that Nutanix can be a CPU and Network/Comm ‘hog’ on host resources with frequent VM motions, use of private and public networks per node and caching/back-up going on in local and remote nodes, etc.  [That overhead is a big reason why Nutanix recently announced their home-built hypervisor, ”Acropolis”.]  And let’s not forget the fact that Nutanix relies only on SSD (flash first) data caching. We all know what happens when SSDs start to reach their fill capacity, more latency with reduced IOPS/bandwidth. Not a good combination. And, unlike ScaleIO, Nutanix doesn’t support true storage tiering to optimize performance and TCO, though they do perform load balancing.

Our customers running ScaleIO in production and mission critical operations are telling us just how much they value ScaleIO’s inherent HA/resilience, elasticity, performance and flexibility to support existing and newly acquired host application servers regardless of their hardware performance, CPU version, DAS capacity, storage media or underlying platform software environment.  And, of course, ScaleIO really scales!  This means they can repurpose their older host servers for creating a single hyper-converged server SAN layer and freely add new, updated host servers (and DAS devices) as the need arises.   ScaleIO is completely HW host server and media device agnostic. This flexibility means you can support a variety of heterogeneous operating systems, hypervisors and CPUs with total hardware configuration asymmetry.

Dynamically adding or removing hosts and storage devices non-disruptively on the fly is another ScaleIO advantage land differentiator from Nutanix. ScaleIO easily scales up to thousands of nodes. Nutanix typically tops off with much smaller node count (i.e., < 50).  Whether you need it or not, with Nutanix you’re going to get both additional compute and storage with each box added. And what about all your existing host servers?  They can’t be used with Nutanix making them instant boat anchors.

More is better.  But we all know that in certain situations “Less is better” too. Here again, ScaleIO is a clear standout.  With a ScaleIO “pure software” SDS system you get less:

  • Less $$/GB (Nutanix nodes aren’t cheap)
  • Less rigid/fixed HW constraints (Nutanix boxes forces new HW with fixed (and limited) storage
  • Less total cost of ownership while delivering better scaling of bandwidth and IOPS
  • Less application and bandwidth bottlenecks through QoS features and intelligent auto-balancing features.

This obviously isn’t the complete list either but supports the adage that with ScaleIO, less is more!  Bottom-line:  no other hyper-converged storage scale-out SAN solution, whether pure software based SDS or HW appliance delivers the sheer performance, scalability, flexibility, robust enterprise features, efficiency and lower TCO than ScaleIO.  Not Nutanix.  Not anybody.  That’s why ScaleIO is ideal for a wide range of environments including the largest Cloud and Enterprise high performance applications.  And, there are good reasons why you don’t see Nutanix much in the enterprise or hyper-scale use cases.  Nutanix’s sweet spot is limited to scaling to support only the needs of mid-sized and mid-tier organizations.

When you compare ScaleIO’s overall design, scalability and performance capabilities (and advantages) you’ll find Nutanix pales in comparison.  Evaluate ScaleIO now with a free and frictionless download. Experience more, and less.

About the Author: Rodger Burkley